diff --git a/libpcsxcore/new_dynarec/linkage_arm.S b/libpcsxcore/new_dynarec/linkage_arm.S
-index bbc52c3..83c5b08 100644
+index 1d8cefa..528929f 100644
--- a/libpcsxcore/new_dynarec/linkage_arm.S
+++ b/libpcsxcore/new_dynarec/linkage_arm.S
-@@ -437,7 +437,7 @@ FUNCTION(cc_interrupt):
+@@ -438,7 +438,7 @@ FUNCTION(cc_interrupt):
str r1, [fp, #LO_pending_exception]
and r2, r2, r10, lsr #17
add r3, fp, #LO_restore_candidate
@@ str r10, [fp, #LO_reg_cop0+36] /* Count */
ldr r4, [r2, r3]
mov r10, lr
-@@ -527,7 +527,7 @@ FUNCTION(new_dyna_leave):
+@@ -528,7 +528,7 @@ FUNCTION(new_dyna_leave):
ldr r0, [fp, #LO_last_count]
add r12, fp, #28
add r10, r0, r10
ldmfd sp!, {r4, r5, r6, r7, r8, r9, sl, fp, ip, pc}
.size new_dyna_leave, .-new_dyna_leave
-@@ -644,7 +644,7 @@ FUNCTION(new_dyna_start):
+@@ -645,7 +645,7 @@ FUNCTION(new_dyna_start):
\readop r0, [r1, r3, lsl #\tab_shift]
.endif
movcc pc, lr
bx r1
.endm
-@@ -679,7 +679,7 @@ FUNCTION(jump_handler_read32):
+@@ -680,7 +680,7 @@ FUNCTION(jump_handler_read32):
mov r0, r1
add r2, r2, r12
push {r2, lr}
blx r3
ldr r0, [fp, #LO_next_interupt]
-@@ -707,7 +707,7 @@ FUNCTION(jump_handler_write_h):
+@@ -708,7 +708,7 @@ FUNCTION(jump_handler_write_h):
add r2, r2, r12
mov r0, r1
push {r2, lr}
ldr r0, [fp, #LO_next_interupt]
diff --git a/libpcsxcore/new_dynarec/linkage_arm64.S b/libpcsxcore/new_dynarec/linkage_arm64.S
-index 698bd78..798abea 100644
+index 7df82b4..79298e4 100644
--- a/libpcsxcore/new_dynarec/linkage_arm64.S
+++ b/libpcsxcore/new_dynarec/linkage_arm64.S
-@@ -119,7 +119,7 @@ FUNCTION(cc_interrupt):
+@@ -123,7 +123,7 @@ FUNCTION(cc_interrupt):
str wzr, [rFP, #LO_pending_exception]
and w2, w2, rCC, lsr #17
add x3, rFP, #LO_restore_candidate
# str rCC, [rFP, #LO_reg_cop0+36] /* Count */
ldr w19, [x3, w2, uxtw]
mov x21, lr
-@@ -227,7 +227,7 @@ FUNCTION(new_dyna_start):
+@@ -231,7 +231,7 @@ FUNCTION(new_dyna_start):
FUNCTION(new_dyna_leave):
ldr w0, [rFP, #LO_last_count]
add rCC, rCC, w0
ldp x19, x20, [sp, #16*1]
ldp x21, x22, [sp, #16*2]
ldp x23, x24, [sp, #16*3]
-@@ -245,7 +245,7 @@ FUNCTION(new_dyna_leave):
+@@ -249,7 +249,7 @@ FUNCTION(new_dyna_leave):
/* w0 = adddr/data, x1 = rhandler, w2 = cycles, x3 = whandler */
ldr w4, [rFP, #LO_last_count]
add w4, w4, w2
.macro memhandler_post
diff --git a/libpcsxcore/new_dynarec/new_dynarec.c b/libpcsxcore/new_dynarec/new_dynarec.c
-index 1452db3..8200e44 100644
+index 2f77516..21481bc 100644
--- a/libpcsxcore/new_dynarec/new_dynarec.c
+++ b/libpcsxcore/new_dynarec/new_dynarec.c
-@@ -44,10 +44,10 @@ static int sceBlock;
- #define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0]))
- #endif
-
--//#define DISASM
--//#define assem_debug printf
-+#define DISASM
-+#define assem_debug printf
- //#define inv_debug printf
--#define assem_debug(...)
-+//#define assem_debug(...)
- #define inv_debug(...)
-
- #ifdef __i386__
-@@ -423,6 +423,9 @@ static int doesnt_expire_soon(void *tcaddr)
+@@ -521,6 +521,9 @@ static int doesnt_expire_soon(void *tcaddr)
// This is called from the recompiled JR/JALR instructions
void noinline *get_addr(u_int vaddr)
{
u_int page=get_page(vaddr);
u_int vpage=get_vpage(vaddr);
struct ll_entry *head;
-@@ -4393,13 +4396,15 @@ void do_cc(int i,signed char i_regmap[],int *adj,int addr,int taken,int invert)
- }
+@@ -4790,13 +4793,15 @@ void do_cc(int i,signed char i_regmap[],int *adj,int addr,int taken,int invert)
+ #endif
emit_addimm_and_set_flags(cycles,HOST_CCREG);
jaddr=out;
- emit_jns(0);
}
add_stub(CC_STUB,jaddr,idle?idle:out,(*adj==0||invert||idle)?0:(count+2),i,addr,taken,0);
}
-@@ -4807,7 +4812,8 @@ static void rjump_assemble(int i,struct regstat *i_regs)
+@@ -5204,7 +5209,8 @@ static void rjump_assemble(int i,struct regstat *i_regs)
// special case for RFE
emit_jmp(0);
else
+ emit_jmp(0);
//load_regs_bt(branch_regs[i].regmap,branch_regs[i].dirty,-1);
#ifdef USE_MINI_HT
- if(rs1[i]==31) {
-@@ -4912,7 +4918,8 @@ static void cjump_assemble(int i,struct regstat *i_regs)
+ if(dops[i].rs1==31) {
+@@ -5309,7 +5315,8 @@ static void cjump_assemble(int i,struct regstat *i_regs)
else if(nop) {
emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
void *jaddr=out;
add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
}
else {
-@@ -5099,7 +5106,8 @@ static void cjump_assemble(int i,struct regstat *i_regs)
+@@ -5496,7 +5503,8 @@ static void cjump_assemble(int i,struct regstat *i_regs)
emit_loadreg(CCREG,HOST_CCREG);
emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG);
void *jaddr=out;
add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
emit_storereg(CCREG,HOST_CCREG);
}
-@@ -5108,7 +5116,8 @@ static void cjump_assemble(int i,struct regstat *i_regs)
+@@ -5505,7 +5513,8 @@ static void cjump_assemble(int i,struct regstat *i_regs)
assert(cc==HOST_CCREG);
emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
void *jaddr=out;
- emit_jns(0);
+ //emit_jns(0);
+ emit_jmp(0);
- add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
+ add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
}
}
-@@ -5210,7 +5219,8 @@ static void sjump_assemble(int i,struct regstat *i_regs)
+@@ -5607,7 +5616,8 @@ static void sjump_assemble(int i,struct regstat *i_regs)
else if(nevertaken) {
emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
void *jaddr=out;
add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
}
else {
-@@ -5366,7 +5376,8 @@ static void sjump_assemble(int i,struct regstat *i_regs)
+@@ -5763,7 +5773,8 @@ static void sjump_assemble(int i,struct regstat *i_regs)
emit_loadreg(CCREG,HOST_CCREG);
emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG);
void *jaddr=out;
add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
emit_storereg(CCREG,HOST_CCREG);
}
-@@ -5375,7 +5386,8 @@ static void sjump_assemble(int i,struct regstat *i_regs)
+@@ -5772,7 +5783,8 @@ static void sjump_assemble(int i,struct regstat *i_regs)
assert(cc==HOST_CCREG);
emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
void *jaddr=out;
- emit_jns(0);
+ //emit_jns(0);
+ emit_jmp(0);
- add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
+ add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
}
}
-@@ -5863,7 +5875,7 @@ void unneeded_registers(int istart,int iend,int r)
+@@ -6211,7 +6223,7 @@ void unneeded_registers(int istart,int iend,int r)
// R0 is always unneeded
u|=1;
// Save it
gte_unneeded[i]=gte_u;
/*
printf("ur (%d,%d) %x: ",istart,iend,start+i*4);
-@@ -8387,6 +8399,7 @@ int new_recompile_block(int addr)
+@@ -8756,6 +8768,7 @@ int new_recompile_block(u_int addr)
// This allocates registers (if possible) one instruction prior
// to use, which can avoid a load-use penalty on certain CPUs.
+#if 0
for(i=0;i<slen-1;i++)
{
- if(!i||(itype[i-1]!=UJUMP&&itype[i-1]!=CJUMP&&itype[i-1]!=SJUMP&&itype[i-1]!=RJUMP))
-@@ -8543,6 +8556,7 @@ int new_recompile_block(int addr)
+ if (!i || !dops[i-1].is_jump)
+@@ -8912,6 +8925,7 @@ int new_recompile_block(u_int addr)
}
}
}
/* Pass 6 - Optimize clean/dirty state */
clean_registers(0,slen-1,1);
-@@ -8842,6 +8856,11 @@ int new_recompile_block(int addr)
+@@ -9217,6 +9231,11 @@ int new_recompile_block(u_int addr)
case SPAN:
pagespan_assemble(i,®s[i]);break;
}
-+
+#ifdef DRC_DBG
-+ if(itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=RJUMP)
-+ wb_dirtys(regs[i].regmap,regs[i].dirty);
++ // write-out non-consts, consts are likely different because of get_final_value()
++ if (!dops[i].is_jump)
++ wb_dirtys(regs[i].regmap,regs[i].dirty&~regs[i].loadedconst);
+#endif
- if(itype[i]==UJUMP||itype[i]==RJUMP||(source[i]>>16)==0x1000)
+ if (dops[i].is_ujump)
literal_pool(1024);
else
-@@ -8950,7 +8969,7 @@ int new_recompile_block(int addr)
- }
- }
- // External Branch Targets (jump_in)
-- if(copy+slen*4>(void *)shadow+sizeof(shadow)) copy=shadow;
-+ if(copy+slen*4>(void *)shadow+sizeof(shadow)) {copy=shadow;printf("shadow overflow\n");}
- for(i=0;i<slen;i++)
- {
- if(bt[i]||i==0)
-@@ -9065,6 +9084,10 @@ int new_recompile_block(int addr)
- }
- expirep=(expirep+1)&65535;
+@@ -9451,6 +9470,10 @@ int new_recompile_block(u_int addr)
}
+ #ifdef ASSEM_PRINT
+ fflush(stdout);
++#endif
+#ifdef DRC_DBG
+printf("new_recompile_block done\n");
+fflush(stdout);
-+#endif
+ #endif
return 0;
}
-